The Hundred-Day Mark

The 2026 Iran war reached its hundredth day on 7 June. The week leading up to it was the most kinetically active period since the April 8 ceasefire. Iranian drones struck Kuwait International Airport on 3 June, killing one person and injuring more than sixty; US forces struck Qeshm Island the same night; Iran targeted Bahrain and a vessel near the Strait of Hormuz. Secretary of State Rubio said on 2 June that the war is "over" - a statement made the same day both sides were exchanging strikes.¹ The 60-day memorandum of understanding that Trump described as "largely negotiated" on 23 May has not been signed. Iran's foreign minister rejected reports of a Hormuz reopening agreement on 3 June, the same day the House passed a war powers resolution 215-208 - the first time either chamber has passed such a measure since the conflict began, with four Republicans joining Democrats.²

The energy backdrop remains the transmission mechanism into digital infrastructure. Brent crude and European natural gas have not normalised since the February 28 campaign began. The Strait of Hormuz remains subject to Iranian control over mine clearance, with the draft MOU requiring Iran to remove mines within 30 days of signature - a signature that has not yet occurred. DC models underwritten before February 28 still carry energy assumptions that do not reflect the current environment.³

Against that backdrop, the week also produced two data points about the scale of the AI infrastructure buildout. Alphabet's $80 billion equity raise and SoftBank's €75 billion France commitment are covered in their own sections below. The juxtaposition is not incidental: the infrastructure required to run AI at scale is now larger than any single balance sheet or corridor can comfortably contain, and the geopolitical environment is actively complicating the corridors that were supposed to carry it.

Iranian drones hit Kuwait International Airport on 3 June, killing one person, injuring more than sixty, damaging two passenger terminals, and forcing a temporary suspension of flights. Iran said it was targeting US bases in Kuwait and Bahrain and a vessel near the Strait of Hormuz; US Central Command said it struck Qeshm Island in response, disabling an Iranian oil tanker heading toward Kharg Island.¹ Joint Chiefs Chairman General Dan Caine described the incidents as "all below the threshold of restarting major combat operations." Defense Secretary Hegseth said the ceasefire "certainly holds."⁵

The diplomatic situation is the more consequential variable for digital infrastructure investors. The draft MOU, as reported by Axios on 24 May, would reopen Hormuz with no tolls, require Iran to clear mines within 30 days, lift the US naval blockade on Iranian ports, and establish a framework for nuclear talks.⁶ Iran's foreign minister rejected reports of a Hormuz reopening agreement on 3 June. The House war powers resolution, while likely to be vetoed if it clears the Senate, signals that domestic political support for continued military operations is eroding - a variable that may affect the administration's negotiating posture and timeline more than the resolution itself.² Trump described the vote as "meaningless" on Truth Social, while simultaneously stating that final negotiations are ongoing.

For PE investors in digital infrastructure: the two scenarios that matter are signature of the MOU within the next two to three weeks, or a return to sustained kinetic exchange. The corridor risk for Gulf and Red Sea cable assets, the energy price environment, and the CFIUS lens on Gulf sovereign co-investment all shift materially depending on which of those resolves first.

On 31 May, the US Department of Commerce's Bureau of Industry and Security issued guidance clarifying that export-licence requirements for advanced AI chips apply to any entity whose parent or headquarters is in China or Macau, regardless of where the purchasing subsidiary is located. The guidance explicitly closes the channel through which Chinese-owned companies operating in Malaysia, Singapore, and other third countries had been acquiring Nvidia Blackwell-class and equivalent chips without the licences that would have been required for direct purchase from inside China.⁸ The guidance is a direct consequence of the May 2025 decision not to enforce the AI Diffusion Rule, which left the subsidiary channel without clear restriction for over a year.

The same week, The New York Times published an analysis by Wirescreen, a Chinese corporate intelligence firm, of more than 3,800 Chinese military and government procurement records. The analysis found that PLA units sought to acquire Nvidia AI chips (including A100, H100, A800, and H800 models) more than 500 times between 2019 and 2025. The procurement records spanned nearly every branch of the Chinese military, including units working on nuclear explosive simulations, offensive cyber operations, and weapons research. The records document procurement attempts and orders rather than confirmed deliveries; the Wirescreen analyst nonetheless described the findings as showing "directly and irrefutably" that US technology had been sought to equip the Chinese military.⁹

The investor implication runs through two channels. The first is compliance: data center assets with Chinese-owned tenants or partners (even where those entities are domiciled outside China) now face a materially tighter regulatory environment than they did seven days ago. Operators that have not mapped ownership chains through to ultimate Chinese-HQ'd beneficiaries need to do so before the next financing event. The second is structural: the BIS guidance accelerates the bifurcation of AI infrastructure stacks. Facilities that had hoped to remain "neutral" (serving both Western and Chinese AI workloads) are finding that the regulatory architecture is closing that option. The addressable customer base for any asset with Chinese-linked tenancy is narrowing.

At the Choose France summit on 31 May, SoftBank Group announced a programme of up to €75 billion to develop and operate 5 GW of AI data center capacity in France, described as the largest announced AI infrastructure commitment in European history. PE investors should note that the €75 billion represents the total project investment value rather than an outright SoftBank balance sheet commitment: SoftBank is positioned as anchor investor and operator, with the programme structured to attract co-investment and project debt alongside its own capital. The first phase targets 3.1 GW by 2031, at an initial cost of €45 billion, across three sites in Hauts-de-France: Dunkirk (Loon-Plage), Bosquel, and Bouchain. EDF is partnering on the Bouchain site, providing the land of a former power plant. Schneider Electric will develop a large-scale manufacturing cluster at the Port of Dunkirk alongside SoftBank's data center operations, producing enclosures and power modules for the buildout.¹⁰

Masayoshi Son's stated rationale deserves attention as a market signal rather than diplomatic courtesy: France's "industrial capabilities, talent base, and national ambition," with nuclear power availability cited explicitly as a structural advantage. This is not a generic commitment to Europe. It is a specific bet on France's combination of low-carbon baseload power, available industrial land, and a government prepared to move on permitting: the same three variables that are constraining AI infrastructure deployment in every other market this series has covered. The EDF partnership at Bouchain is the concrete expression of the nuclear power thesis: a former power plant site with existing grid connection, supplied by a state-owned utility with a domestic mandate to support the project.

The read-through for PE investors in European digital infrastructure is twofold. First, the SoftBank commitment will pull third-party suppliers, financing, and talent toward France in ways that alter the relative attractiveness of competing European DC markets over the next three to five years. Second, the project's structure (nuclear anchor, state utility partner, manufacturing co-location, government fast-track permitting) represents a template for how sovereign AI infrastructure deals may be structured in Europe going forward. Assets that can replicate elements of that structure in other EU markets may command a premium; assets dependent on grid-connected market-rate power in the same timeframe face a more difficult competitive position.

On 1 June, Alphabet announced an $80 billion equity raise across three tranches: $30 billion in underwritten public offerings, $40 billion through an at-the-market programme beginning in Q3, and a $10 billion private placement to Berkshire Hathaway (the deal was subsequently priced at $84.75 billion after strong investor demand).⁴ The stated reason was direct: "The company is experiencing strong demand for its AI solutions and services from enterprises and consumers, at levels that are exceeding the company's available supply." Google Cloud revenue grew 63% year-on-year in Q1 2026; the Cloud backlog nearly doubled to over $460 billion. Alphabet was already the first tech company since Motorola in 1997 to issue a 100-year bond; this equity raise is the logical extension of the same argument.

This tile earns its editorial slot not because of the dollar amount (large capital markets transactions by public companies are routine) but because of what the transaction structure reveals about the state of AI infrastructure demand. Alphabet's internal cash generation is among the highest of any company in history. The fact that it has concluded this is insufficient to fund its buildout at the pace demand requires is the signal. For PE investors in independent data center capacity: when the largest hyperscalers cannot build fast enough from their own resources, the structural case for third-party, independently operated capacity becomes materially stronger. The Berkshire Hathaway placement (Greg Abel's first significant AI infrastructure bet, committing roughly $31 billion to Alphabet across recent quarters) adds a further signal: value capital institutions that have historically avoided growth-multiple technology are now treating AI compute capacity as a long-duration infrastructure asset class.

• 01
  For any data center asset with tenants or partners whose ultimate parent is headquartered in China or Macau: has the ownership chain been fully mapped against the 31 May BIS guidance? Does any tenant relationship require a licence that was not previously required?

  BIS guidance on Chinese-subsidiary AI chip exports - Commerce Department, 31 May 2026
• 02
  For Gulf or Red Sea corridor cable assets: at what point does extended Hormuz disruption trigger SLA penalties, and what is the revenue break-even between a 30-day and a 90-day corridor outage? The MOU stall this week removes the near-term reopening assumption; the force majeure clause that was adequate under a ceasefire-holding scenario may not be adequate under the current exchange-of-strikes scenario.

  Kuwait airport strike; MOU stall - Washington Post / Axios, 3-4 June 2026
• 03
  For European data center assets in or near France: does the SoftBank 3.1 GW buildout in Hauts-de-France between now and 2031 represent a competitive threat or a demand catalyst for this asset? The answer depends on two specifics: whether the asset sits within the primary hyperscale tenant catchment of the northern France cluster, and whether its power supply is grid-connected at market rates (in which case it competes against nuclear-anchored capacity it cannot match on price) or contracted and captive (in which case it may benefit from connectivity and supply-chain proximity).

  SoftBank €75bn France commitment - DCD / Euronews, 31 May 2026

• 04
  For all data center assets carrying energy cost assumptions set before February 28: what is the IRR impact across the remaining hold period at current Brent and European gas prices? The MOU stall means no near-term normalisation of Hormuz-related energy costs; any valuation or financing event in the next 90 days should carry a current-pricing scenario alongside the original underwriting model.

  Iran war energy shock; MOU stall - ongoing
• 05
  For US data center assets in Wisconsin, North Carolina, Oklahoma, Texas, Virginia, or Ohio: what is the IRR delta between the underwriting model's power cost assumptions and a full-cost-of-service tariff equivalent to the Wisconsin We Energies precedent? Oklahoma's act takes effect 1 July 2026; the question for assets in the three remaining states is not whether this template arrives, but whether the hold period has enough runway to exit before it does.

  Oklahoma Data Center Consumer Ratepayer Protection Act - signed mid-May 2026, effective 1 July 2026
• 06
  For assets with captive nuclear power in the investment thesis: the DOE reactor pilot July 4 deadline is 30 days away. What is the asset's dependency on a specific reactor project achieving criticality on that timeline, and what is the fallback if the deadline slips?

  DOE Reactor Pilot Programme - July 4, 2026 criticality deadline

• 07
  If the ownership chain includes any entity within scope of the BIS 50% Rule effective 10 November 2026: what is the remediation path, and does the timeline risk to closing or refinancing change materially if external counsel identifies a breach after Q3 2026? The 31 May guidance signals a more active BIS posture; an operator that discovers a scoping issue in October has approximately six weeks to remediate before the enforcement date.

  BIS 50% Rule enforcement deadline; 31 May guidance enforcement signal
• 08
  For fund structures with Gulf sovereign co-investors in US AI-adjacent assets where no CFIUS voluntary notice has been filed: has legal counsel assessed whether the 3 June escalation changes the risk calculus? A CFIUS reviewer evaluating a Gulf sovereign co-investment this week is operating in a materially different environment than Q4 2025; the question is not whether the structure was clean at inception but whether the home-country conduct of the co-investor since April 8 introduces grounds for a mandatory referral or a contested review.

  Iran war Day 96 escalation; CFIUS elevated risk - ongoing

• 09
  For European AI infrastructure assets seeking sovereign or government-backed capital: which single element of the SoftBank France structure is the binding constraint for this asset? If the answer is nuclear-anchored power supply, the asset's grid dependency needs to be addressed before approaching sovereign LPs. If it is fast-track permitting, the question is whether a government offtake or co-investment commitment can be structured to unlock equivalent regulatory treatment. The SoftBank template is not replicable wholesale; identifying which element is out of reach determines whether a partial structure is worth pursuing at all.

  SoftBank / EDF / Schneider Electric France commitment - 31 May 2026

• Map ownership chains for all assets with Chinese-linked tenants or partners against the 31 May BIS guidance. The compliance landscape changed this week. Any asset that has not audited whether its tenant or partner relationships involve Chinese-HQ'd entities operating through overseas subsidiaries now carries an unquantified compliance risk. This is material for any financing event or exit process in the next 90 days.
• Re-run Gulf corridor scenario analysis in light of the 3 June escalation and MOU stall. The Kuwait airport strike and Qeshm response represent a qualitative change in the risk environment for Gulf and Red Sea assets. Any portfolio company with SLA dependency on those corridors may benefit from a contingency assessment before the next deal committee or valuation event.
• For European assets: assess proximity and connectivity to the SoftBank France buildout in Hauts-de-France. The €75 billion commitment will reshape northern French digital infrastructure over the next three to five years. Assets within the catchment of that buildout - in terms of talent, connectivity, and supply chain - may benefit; assets competing for the same market may face pressure.

• Initiate BIS 50% Rule compliance mapping before Q3 2026. The 31 May enforcement action signals a more active BIS posture. November 10 is five months away. Operators with any PRC-adjacent counterparties need external counsel engaged well before the deadline, given the complexity of the ownership-chain mapping required and the demonstrated willingness to act.
• For AI infrastructure assets in development: model the Alphabet equity raise as a demand-side reference point. A company generating $174 billion in annual operating cash flow concluded it cannot fund its AI buildout from internal resources. That demand signal has direct implications for the underwriting assumptions behind independent third-party capacity: committed anchor tenants and pre-contracted capacity are now more defensible than speculative development.

• Develop captive power strategies for data center holdings in markets where grid cost reallocation is becoming a pattern. Wisconsin, North Carolina, and Oklahoma have all acted. Texas, Virginia, and Ohio are the next markets in range. The SoftBank France nuclear anchor confirms that captive low-carbon baseload is becoming a structural differentiator at sovereign investment scale. Assets with dedicated generation or fixed PPAs appear likely to command a durable exit multiple premium over grid-dependent peers as this direction becomes entrenched.